aidsorb.transforms

This module provides helper functions and classes for transforming point clouds.

Note

The pcd must be a ndarray and have shape of (N, 3+C).

Tip

For implementing your own transforms, have a look at the transforms tutorial. For more flexibility, consider implementing them as callable instances of classes.

class aidsorb.transforms.Center

Bases: object

Center the coordinates of a point cloud by subtracting their centroid.

Examples

>>> pcd = np.array([[1., 2., 3., 5.], [2., 4., 5., 3.]])
>>> center = Center()
>>> center(pcd)
array([[-0.5, -1. , -1. ,  5. ],
       [ 0.5,  1. ,  1. ,  3. ]])

>>> center(pcd).mean(axis=0)
array([0., 0., 0., 4.])

class aidsorb.transforms.Identity

Bases: object

Leave the point cloud unchanged.

Examples

>>> pcd = np.random.randn(300, 4)
>>> identity = Identity()
>>> np.array_equal(identity(pcd), pcd)
True

class aidsorb.transforms.Jitter(std=0.01)

Bases: object

Jitter the coordinates of a point cloud by adding normal noise.

Parameters:

std (float, default=0.01) – The standard deviation of the normal noise.

Examples

>>> pcd = np.random.randn(100, 5)
>>> jitter = Jitter()
>>> new_pcd = jitter(pcd)
>>> new_pcd.shape
(100, 5)

>>> from aidsorb.utils import split_pcd
>>> coords, feats = split_pcd(pcd)
>>> new_coords, new_feats = split_pcd(new_pcd)

>>> np.array_equal(new_coords, coords)  # Coordinates are affected.
False
>>> np.array_equal(new_feats, feats)  # Features are not affected.
True

class aidsorb.transforms.RandomErase(n_points=5)

Bases: object

Randomly erase a number of points from the point cloud.

Todo

Consider adding the option for a fraction of points to be erased.

Parameters:

n_points (int, default=5) – Number of points to be erased.

Examples

>>> pcd = np.random.randn(100, 5)
>>> erase = RandomErase(n_points=10)
>>> erase(pcd).shape
(90, 5)

class aidsorb.transforms.RandomRotation

Bases: object

Randomly rotate the coordinates of a point cloud.

Examples

>>> pcd = np.random.randn(25, 4)
>>> rot = RandomRotation()
>>> new_pcd = rot(pcd)
>>> new_pcd.shape
(25, 4)

>>> from aidsorb.utils import split_pcd
>>> coords, feats = split_pcd(pcd)
>>> new_coords, new_feats = split_pcd(new_pcd)

>>> np.array_equal(new_coords, coords)  # Coordinates are affected.
False
>>> np.array_equal(new_feats, feats)  # Features are not affected.
True

aidsorb.transforms.transform_pcd(pcd, tfm)

Transform the coordinates of a point cloud.

For molecular point clouds, only rigid transformations are recommended.

Parameters:

• pcd (array of shape (N, 3+C)) – The original point cloud.

• tfm (array of shape (3, 3)) – The transformation matrix.

Returns:

new_pcd – The transformed point cloud.

Return type:

array of shape (N, 3+C)

Raises:

ValueError – If pcd or tfm do not have the expected shape.

Examples

>>> pcd = np.array([[3, -9, 2, 6], [3, 4, -1, 8]])
>>> tfm = np.array([[0, -1, 0], [1, 0, 0], [0, 0, 1]])
>>> transform_pcd(pcd, tfm)
array([[ 9,  3,  2,  6],
       [-4,  3, -1,  8]])

>>> pcd = np.random.randn(424, 2)  # Invalid shape.
>>> transform_pcd(pcd, tfm)
Traceback (most recent call last):
    ...
ValueError: Expecting array of shape (N, 3+C) but got array of shape (424, 2)!